BACKGROUND: Hormone profile evaluation traditionally requires an expert review of results for adequate laboratory section test release. Analysis of these results using Boolean rules is an interesting alternative to reduce the number of results that require manual review. MATERIAL AND METHOD: We evaluated the utilization of Boolean rules using Instrument Manager software and Architect analyzer, mainly performing sex and thyroid hormones measurement. The intervention was evaluated on: a) number of rules and its easiness of construction; b) blind comparison of results evaluation by clinical pathologist (printed results) and set of rules in 940 consecutive tests. RESULTS: Rule creation was a complex and arduous task, especially due to hormonal profiles with several different request patterns. It was necessary to use a set of 153 Boolean (if…then) rules, in a specific order. This set of rules agreed with expert opinion in 97.9% (920 tests). Rules hold 25 tests (2.7%) and the clinical pathologist only nine tests. There was discordance in 20 cases; rules did not hold only two cases: a beta-hCG in a male patient (that prompted the creation of a new rule) and a complete thyroid profile lacking only TSH request (pathologist opted to review the original request). CONCLUSION: Creation of an efficient set of Boolean rules proved to be a complex task requiring both technical and logics knowledge, but allowing optimization of laboratory workload. We achieved excellent concordance between the set of rules and clinical pathologist manual review, in a safe, fast and low cost system.
Logic; Hormones; Software validation; Artificial intelligence
